Humans can be easily recognized because of their uniqueness in nature, also with their demographic characteristics namely gender, ethnicity such as race, and age correspondingly. In the present decades, numerous numbers of studies have been conducted in the areas such as biological, psychological and cognitive sciences in order to identify how the human brain can memorize, perceives and characterizes human faces. In addition, specific computational developments are performed to attain numerous aspects of this problem. This work attempts to present a novel race detection technique by exploiting the face shape features. Here, the developed model comprises two stages such as feature extraction and detection. In the primary phase, the feature extraction is performed, here the face shape, and face color-based feature, is mined. Particularly, Speeded-up Robust Transform (SURF), and Maximally Stable Extremal Regions (MSER), is extracted in shape features and dense color features are extracted as a color feature. In large dimensions, extracted features are presented; they are changed in the Principle Component Analysis (PCA) technique that is considered the strongest technique to solve the “curse of dimensionality”. Subsequently, dimensional minimized features are fed to Deep Belief Network (DBN), whereas the race is identified. Furthermore, regarding the prediction to make the developed model highly efficient, the DBN weight is fine-tuned with a novel technique called mutated Salp Swarm Optimization Algorithm. The adopted technique is evaluated with existing techniques regarding the accuracy and error performance.
Buy article PDF